To the Editor:
The World Health Organization has estimated that more than two billion persons in the world carry latent tuberculosis infection (LTBI). The diagnosis and treatment of LTBI could have an important impact in preventing the development of active tuberculosis (TB), a disease that causes 1.4 million deaths annually [1]. The current standard treatment of LTBI is nine months of isoniazid (INH) [2]. To improve adherence, a shorter regimen of 4 months rifampicin (RIF) is being evaluated in a multicentre randomised trial (CIHR MCT-94831, registered at www.controlled-trials.com/ISRCTN05675547). This ongoing trial offers an opportunity to study potential biomarkers as surrogates of successful prevention of active disease [3].
Several changes in lipid metabolism could potentially be utilised as biomarkers of effectiveness of LTBI treatment with RIF or INH. RIF is an important stimulator of the pregnane X receptor (PXR), which has been hypothesised to increase the blood levels of high density lipoprotein cholesterol (HDLC) and its protein component apolipoprotein A-1 (apoA) [4, 5]. This hypothesis was supported by observational studies that demonstrated an increased plasma level of HDLC among epileptic patients who were taking anticonvulsant medications which activate the PXR [6]. By contrast, treatment with INH was associated with lower cholesterol blood levels in one small study [7]. ApoA and apolipoprotein B (apoB), a protein component of low density lipoprotein cholesterol (LDLC), have stable serum blood levels without post-prandial changes and low within-individual variability [8]. Total cholesterol and HDLC are inexpensive to measure, reliably measured using non-fasting blood samples, and have lower within-individual variability than triglycerides or LDLC [9].
The objective of this study was to perform a preliminary assessment of total cholesterol, HDLC, apoA and apoB as potential biomarkers, by comparing levels of these four substances before treatment and after one month of treatment with RIF or INH.
From participants in the ongoing multicentre trial of LTBI treatment (CIHR MCT-94831), 15 randomised to 4 months RIF were selected randomly, and 15 randomised to 9 months INH, matched on age (within 2 years) and sex, were selected. LTBI was defined as a positive tuberculin skin test and/or a positive interferon-γ release assay (positive as defined in Canadian TB standards [10]). To limit the variability of metabolic conditions between subjects, we restricted participants to adults between the age of 18 and 45 years. Subjects were excluded if they were consuming alcohol daily or taking other drugs that could induce cytochrome P450 (carbamazepine, phenytoin, phenobarbital, felbamate, topiramate, lamotrigine, griseofulvin, nevirapine and oral contraceptives), as these may influence the effect of RIF on lipid blood levels.
After signing informed consent, participants provided serum samples before treatment and after 1 month of self-administered LTBI treatment. Subjects were not fasting at the time of blood sampling since normal food consumption has no effect on apoA and apoB and a clinically unimportant effect on total cholesterol and HDLC [8]. Serum samples were labelled with study identity numbers and stored at -80°C until assays of total cholesterol, HDLC, apoA and apoB were performed. The identity of participants and study drug taken was known only to one investigator (C. Valiquette); all other investigators remained blinded to study drug for data analysis and interpretation. This study was approved by a research ethics board of the McGill University Health Centre (file number 11-046-SDR).
Treatment effect on lipids was assessed by comparing the post-treatment serum lipids to the pretreatment levels in each arm, using a paired t-test. The mean change in lipid (treatment effect) was compared between the two intervention groups using linear regression statistics. Residual confounding by age and sex, and potential confounding by other variables (body mass index, alcohol consumption and pretreatment lipid levels) were assessed by comparing different regression models, using stepwise backward regression to select the most parsimonious models. A sample size of 30 subjects was considered adequate for detecting a 10–20% change in serum lipids, considering biological variability of 5–20% [9]. Statistical analyses were conducted using STATA v12.1 (StataCorp LP, College Station, TX, USA).
Pretreatment characteristics of the selected subjects receiving INH or RIF were similar. Except for two subjects, all participants had taken more than 80% of their treatment doses during the first month.
INH treatment was associated with a significant decrease in both total cholesterol and apoB levels (mean changes -0.25 g·L−1 (95% CI -0.45– -0.06) and -0.07 g·L−1 (95% CI -0.12– -0.03), respectively) (fig. 1). The change in serum level of apoB, but not total cholesterol, was significantly different between the two treatment groups, mean difference -0.1 g·L−1 (95% CI -0.2– -0.004).
RIF treatment, by contrast, was associated with an increase in mean apoA levels (fig. 1), although this change was not statistically significant (mean change 0.06 g·L−1 (95% CI -0.01–0.13)).
These preliminary observations suggest that lipid metabolism may be altered significantly by LTBI treatment, particularly within one month of starting INH. These observations require confirmation in a larger study, but offer the promise of novel biomarkers in LTBI treatment.
This study had several important limitations. First the sample size was small; this limited our ability to identify associations between total cholesterol and INH treatment, or between apoA and RIF treatment. We were unable to evaluate potential drug effects on serum triglycerides and LDLC as we had to rely on non-fasting serum samples. However, measurement of lipids that require 12 h of fasting before drawing samples are not very practical for therapeutic monitoring in clinical practice. In addition, we compared the effect of two different modalities of treatment, but did not include an untreated control group with LTBI to assess the spontaneous changes in these lipid levels, although the finding that RIF did not have a significant effect on apoB serum level suggests that INH was responsible for the changes seen. Finally, we could not exclude confounding by unknown factors. However, since the initial assignment to treatment arms in this patient population was random, confounding by unknown factors should have been unlikely.
Due to lack of clinical symptoms or methods of mycobacterial isolation among patients with LTBI, there is currently no way to evaluate patient response to LTBI treatment. Our finding of a significant association between apoB and INH treatment suggests that changes in serum apoB may be a surrogate of adherence to, and possibly also effectiveness of, INH therapy. However, this requires further evaluation with a larger number of subjects to correlate these changes with patient adherence and effectiveness of INH. In addition, this association, which, to our knowledge, has not been investigated before, may explore a novel mechanism for controlling lipid disorders. Although our observed effect of INH treatment, in reducing apoB level by 8.4%, is less than the effect of statins (at least 20%) and other lipid lowering agents (at least 10%), this effect was observed in healthy individuals with normal baseline apoB levels who are different from those treated with lipid lowering agents.
We conclude that apoB may be a potentially useful biomarker for therapeutic monitoring of LTBI treatment with INH; however, further studies with a larger number of patients, treated for longer periods of time and compared with untreated controls, are required.
Acknowledgments
The authors thank Norma Tink (Respiratory Epidemiology and Clinical Research Unit, Montreal Chest Institute, McGill University, Montreal, QC, Canada) for her assistance during the study.
Footnotes
Support statement: This work was supported by CIHR (CIHR MCT-94831). A.S. Albanna received salary support from the Ministry of Higher Education, Saudi Arabia.
Conflict of interest: Disclosures can be found alongside the online version of this article at www.erj.ersjournals.com.
- Received April 11, 2013.
- Accepted April 19, 2013.
- ©ERS 2013